Added initial qemu-eviction and qemu-pagestep

Qemu-eviction seems to get stuck somewhere, potentially in the VC-handler(?) since we use the active gfn after resuming execution. Added qemu-pagestep to show viability of page-stepping for later use.

1 files changed, 63 insertions, 0 deletions

diff --git a/test/qemu-pagestep.c b/test/qemu-pagestep.c
new file mode 100644
index 0000000..4e54161
--- /dev/null
+++ b/test/qemu-pagestep.c
@@ -0,0 +1,63 @@
+#include "test/kvm-eviction.h"
+#include "test/kvm.h"
+#include "test/util.h"
+#include "cachepc/uapi.h"
+
+#include <sys/ioctl.h>
+#include <sys/mman.h>
+#include <signal.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <errno.h>
+#include <err.h>
+#include <string.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+
+void
+monitor(void)
+{
+	struct cpc_event event;
+	int ret;
+
+	ret = ioctl(kvm_dev, KVM_CPC_POLL_EVENT, &event);
+	if (ret && errno == EAGAIN) return;
+	if (ret) err(1, "KVM_CPC_POLL_EVENT");
+
+	if (event.type != CPC_EVENT_TRACK_PAGE)
+		errx(1, "unexpected event type %i", event.type);
+
+	printf("Event: rip:%016llx prev:%08llx next:%08llx ret:%llu\n",
+		vm_get_rip(), event.page.inst_gfn_prev,
+		event.page.inst_gfn, event.page.retinst);
+	printf("\n");
+
+	ret = ioctl(kvm_dev, KVM_CPC_ACK_EVENT, &event.id);
+	if (ret) err(1, "KVM_CPC_ACK_EVENT");
+}
+
+int
+main(int argc, const char **argv)
+{
+	uint32_t arg;
+	int ret;
+
+	setvbuf(stdout, NULL, _IONBF, 0);
+
+	kvm_setup_init();
+
+	pin_process(0, SECONDARY_CORE, true);
+
+	ret = ioctl(kvm_dev, KVM_CPC_RESET);
+	if (ret) err(1, "KVM_CPC_RESET");
+
+	arg = CPC_TRACK_PAGES_RESOLVE;
+	ret = ioctl(kvm_dev, KVM_CPC_TRACK_MODE, &arg);
+	if (ret) err(1, "KVM_CPC_TRACK_MODE");
+
+	while (1) monitor();
+
+	kvm_setup_deinit();
+}
+